This pull request moves swaths of Cascadia to use `til::color` for color
interop. There are still some places where we use `COLORREF`, such as in
the ABI boundaries between WinRT components.
I've also added two more til::color helpers - `with_alpha`, which takes
an existing color and sets its alpha component, and a
`Windows::UI::Color` convertor pair.
Future direction might include a `TerminalSettings::Color` type at the
idl boundary so we can finally stop using UInt32s (!) for color.
## Validation Steps Performed
Tested certain fragile areas:
* [x] setting the background with OSC 11
* [x] setting the background when acrylic is in use (which requires
low-alpha)
It turns out that we weren't really adequately guarding calls to
SetSelectionEnd and friends.
We're clearing the active selection when the window resizes, but we're
doing so by nulling out the std::optional<Selection> it lives in. Later,
though, when we set the selection endpoint we're using "_selection->".
Optional's operator-> has undefined behavior when the optional doesn't
have a value in it.
In our case, it looks like it was returning whatever the value was prior
to it being emptied out. PivotSelection would attempt to access an
out-of-bounds coordinate when the buffer got smaller during a resize.
The solution is to guard both levels of selection endpoint manipulation
in a check for an active selection.
Apparently, this accounts for somewhere between 7% and 14% of our
crashes on 1.0RC1.
Repro was:
Use Win+Arrow to snap the window while in the middle of a selection.
## Summary of the Pull Request
Adds user settings to adjust rendering behavior to mitigate blurry text on some devices.
## References
- #778 introduced this, almost certainly.
## PR Checklist
* [x] Closes#5759, mostly
* [x] I work here.
* [ ] We need community verification that this will help.
* [x] Updated schema and schema doc.
* [x] Am core contributor. Discussed in Monday sync meeting and w/ @DHowett-MSFT.
## Detailed Description of the Pull Request / Additional comments
When we switched from full-screen repaints to incremental rendering, it seems like we exposed a situation where some display drivers and hardware combinations do not handle scroll and/or dirty regions (from `IDXGISwapChain::Present1`) without blurring the data from the previous frame. As we're really close to ship, I'm offering two options to let people in this situation escape it on their own. We hope in the future to figure out what's actually going on here and mitigate it further in software, but until then, these escape hatches are available.
1. `experimental.rendering.forceFullRepaint` - This one restores the pre-778 behavior to the Terminal. On every single frame paint, we'll invalidate the entire screen and repaint it.
2. `experimental.rendering.software` - This one uses the software WARP renderer instead of using the hardware and display driver directly. The theory is that this will sidestep any driver bugs or hardware variations.
One, the other, or both of these may be field-applied by users who are experiencing this behavior.
Reverting #778 completely would also resolve this, but it would give back our largest performance win in the whole Terminal project. We don't believe that's acceptable when seemingly a majority of the users are experiencing the performance benefit with no detriment to graphical display.
## Validation Steps Performed
- [x] Flipped them on and verified with the debugger that they are being applied to the rendering pipeline
- [ ] Gave a private copy to community members in #5759 and had them try whether one, the other, or both resolved their issue.
## Summary of the Pull Request
Based on the discussion in #5479, it seems that the crash is caused by a race condition due to not obtaining the write lock before calling `TriggerFontChange`.
I'm not totally sure if my approach is the right one, but I've taken the lock out of `_RefreshSize` since it seems like all calls of `_RefreshSize` come after a `TriggerFontChange`/`UpdateFont`. Then I just
made sure all calls of `TriggerFontChange`/`UpdateFont` are preceded with a `LockForWriting`.
## PR Checklist
* [x] Closes#5479
* [x] CLA signed.
* [x] Tests added/passed
## Validation Steps Performed
Scrolling to change my font size does not kill the Terminal anymore! 🙌
Web apps apparently will paste the <title> as plaintext before the
actual HTML content from the clipboard. Since this seems to be
widespread behavior across web apps, this isn't just a bug in _some
app_, this is a bug on us. We shouldn't emit the title.
This PR removes the title tag from the generated HTML.
Closes#5347
## Summary of the Pull Request
Before this, if the Search Box was open, new selections would not notify automation clients. This was because the UiaEngine (responsible for notifying automation clients) would remain disabled. Now we're enabling it before the early exit in TermControl's FocusHandler
## PR Checklist
* [X] Will close issue #5421 upon verification
## Validation Steps Performed
Verified using NVDA.
Narrator's behavior is not impacted, for some reason.
Takes the lock inside two routines in `TermControl` that were changing
the selection endpoint while a rendering frame was still drawing,
resulting in several variants of graphical glitches from double-struck
selection boxes to duplicated line text.
## References
- Introduced with #5185
## PR Checklist
* [x] Closes#5471
* [x] Already signed life away to company.
* [x] Manual tests passed since it's visual.
* [x] No extra doc besides the comments.
* [x] Am core contributor: Roar.
The renderer base and specific renderer engine do a lot of work to
remember the previous selection and compensate for scrolling regions and
deltas between frames. However, all that work doesn't quite match up
when the endpoints are changed out from under it. Unfortunately,
`TermControl` doesn't have a robust history of locking correctly in step
with the renderer nor does the renderer's `IRenderData` currently
provide any way of 'snapping' state at the beginning of a frame so it
could work without a full lock. So the solution for now is for the
methods that scroll the display in `TermControl` to take the lock that
is shared with the renderer's frame painter so they can't change out of
sync.
## Validation Steps Performed
- Opened terminal with Powershell core.
Did ls a bunch of times.
Clicked to make selection and held mouse button while wheeling around.
- Opened terminal with Powershell core.;
Did ls a bunch of times.
Clicked to make selection and dragged mouse outside the window to make
auto scroll happen.
- Opened terminal with Powershell core.
Did ls a bunch of times.
Clicked to make selection and released. Wheeled around like a crazy
person to make sure I didn't regress that.
## Summary of the Pull Request
- Adjusts scaling practices in `DxEngine` (and related scaling practices in `TerminalControl`) for pixel-perfect row baselines and spacing at High DPI such that differential row-by-row rendering can be applied at High DPI.
## References
- #5185
## PR Checklist
* [x] Closes#5320, closes#3515, closes#1064
* [x] I work here.
* [x] Manually tested.
* [x] No doc.
* [x] Am core contributor. Also discussed with some of them already via Teams.
## Detailed Description of the Pull Request / Additional comments
**WAS:**
- We were using implicit DPI scaling on the `ID2D1RenderTarget` and running all of our processing in DIPs (Device-Independent Pixels). That's all well and good for getting things bootstrapped quickly, but it leaves the actual scaling of the draw commands up to the discretion of the rendering target.
- When we don't get to explicitly choose exactly how many pixels tall/wide and our X/Y placement perfectly, the nature of floating point multiplication and division required to do the presentation can cause us to drift off slightly out of our control depending on what the final display resolution actually is.
- Differential drawing cannot work unless we can know the exact integer pixels that need to be copied/moved/preserved/replaced between frames to give to the `IDXGISwapChain1::Present1` method. If things spill into fractional pixels or the sizes of rows/columns vary as they are rounded up and down implicitly, then we cannot do the differential rendering.
**NOW:**
- When deciding on a font, the `DxEngine` will take the scale factor into account and adjust the proposed height of the requested font. Then the remainder of the existing code that adjusts the baseline and integer-ifies each character cell will run naturally from there. That code already works correctly to align the height at normal DPI and scale out the font heights and advances to take an exact integer of pixels.
- `TermControl` has to use the scale now, in some places, and stop scaling in other places. This has to do with how the target's nature used to be implicit and is now explicit. For instance, determining where the cursor click hits must be scaled now. And determining the pixel size of the display canvas must no longer be scaled.
- `DxEngine` will no longer attempt to scale the invalid regions per my attempts in #5185 because the cell size is scaled. So it should work the same as at 96 DPI.
- The block is removed from the `DxEngine` that was causing a full invalidate on every frame at High DPI.
- A TODO was removed from `TermControl` that was invalidating everything when the DPI changed because the underlying renderer will already do that.
## Validation Steps Performed
* [x] Check at 150% DPI. Print text, scroll text down and up, do selection.
* [x] Check at 100% DPI. Print text, scroll text down and up, do selection.
* [x] Span two different DPI monitors and drag between them.
* [x] Giant pile of tests in https://github.com/microsoft/terminal/pull/5345#issuecomment-614127648
Co-authored-by: Dustin Howett <duhowett@microsoft.com>
Co-authored-by: Mike Griese <migrie@microsoft.com>
This fixes an issue where a shift+click selection with `copyOnSelect`
enabled would result in copying the content as a single line.
## Detailed Description of the Pull Request / Additional comments
I've been thinking a lot about this issue and how it relates to the
copy/paste discussions we've been having over the past few weeks.
Considering that the majority of users want regular copy, it makes sense
to default to that in this case too.
If a user wants to perform a special form of copy, it makes sense that
they should use their custom keybinding to accomplish that. This kind of
behavior aligns with that kind of philosophy.
## Validation Steps Performed
The following scenarios were tested with `copyOnSelect` enabled.
| scenario | behavior |
|-----------------------------------------|-------------------------------|
| Perform a shift+click selection | content copied w/ newlines |
| right-click | clipboard paste |
| copy keybinding (`singleLine` disabled) | content copied w/ newlines |
| copy keybinding (`singleLine` enabled) | content copied as single line |
Closes#4737
## Summary of the Pull Request
Renderer: Add support for backoff and auto-disable on failed retry
This commit introduces a backoff (150ms * number of tries) to the
renderer's retry logic (introduced in #2830). It also changes the
FAIL_FAST to a less globally-harmful render thread disable, so that we
stop blowing up any application hosting a terminal when the graphics
driver goes away.
In addition, it adds a callback that a Renderer consumer can use to
determine when the renderer _has_ failed, and a public method to kick it
back into life.
Fixes#5340.
This PR also wires up TermControl so that it shows some UI when the renderer tastes clay.
## PR Checklist
* [x] Closes#5340
* [x] cla
* [ ] Tests added/passed
* [ ] Requires documentation to be updated
* [x] I've discussed this with core contributors already.
## Validation Steps Performed
I tested this by dropping the number of retries to 1 and forcing a TDR while doing `wsl cmatrix -u0`. It picked up exactly where it left off.
As a bonus, you can actually still type into the terminal when it's graphically suspended (and `exit` still works.). The block is _entirely graphical_.
## Summary of the Pull Request
Adjusts DirectX renderer to use `til::bitmap` to track invalidation
regions. Uses special modification to invalidate a row-at-a-time to
ensure ligatures and NxM glyphs continue to work.
## References
Likely helps #1064
## PR Checklist
* [x] Closes#778
* [x] I work here.
* [x] Manual testing performed. See Performance traces in #778.
* [x] Automated tests for `til` changes.
* [x] Am core contributor. And discussed with @DHowett-MSFT.
## Detailed Description of the Pull Request / Additional comments
- Applies `til::bitmap` as the new invalidation scheme inside the
DirectX renderer and updates all entrypoints for collecting
invalidation data to coalesce into this structure.
- Semi-permanently routes all invalidations through a helper method
`_InvalidateRectangle` that will expand any invalidation to cover the
entire line. This ensures that ligatures and NxM glyphs will continue
to render appropriately while still allowing us to dramatically reduce
the number of lines drawn overall. In the future, we may come up with
a tighter solution than line-by-line invalidation and can modify this
helper method appropriately at that later date to further scope the
invalid region.
- Ensures that the `experimental.retroTerminalEffects` feature continues
to invalidate the entire display on start of frame as the shader is
applied at the end of the frame composition and will stack on itself
in an amusing fashion when we only redraw part of the display.
- Moves many member variables inside the DirectX renderer into the new
`til::size`, `til::point`, and `til::rectangle` methods to facilitate
easier management and mathematical operations. Consequently adds
`try/catch` blocks around many of the already-existing `noexcept`
methods to deal with mathematical or casting failures now detected by
using the support classes.
- Corrects `TerminalCore` redraw triggers to appropriately communicate
scrolling circumstances to the renderer so it can optimize the draw
regions appropriately.
- Fixes an issue in the base `Renderer` that was causing overlapping
scroll regions due to behavior of `Viewport::TrimToViewport` modifying
the local. This fix is "good enough" for now and should go away when
`Viewport` is fully migrated to `til::rectangle`.
- Adds multiplication and division operators to `til::rectangle` and
supporting tests. These operates will help scale back and forth
between a cell-based rectangle and a pixel-based rectangle. They take
special care to ensure that a pixel rectangle being divided downward
back to cells will expand (with the ceiling division methods) to cover
a full cell when even one pixel inside the cell is touched (as is how
a redraw would have to occur).
- Blocks off trace logging of invalid regions if no one is listening to
optimize performance.
- Restores full usage of `IDXGISwapChain1::Present1` to accurately and
fully communicate dirty and scroll regions to the underlying DirectX
framework. This additional information allows the framework to
optimize drawing between frames by eliminating data transfer of
regions that aren't modified and shuffling frames in place. See
[Remarks](https://docs.microsoft.com/en-us/windows/win32/api/dxgi1_2/nf-dxgi1_2-idxgiswapchain1-present1#remarks)
for more details.
- Updates `til::bitmap` set methods to use more optimized versions of
the setters on the `dynamic_bitset<>` that can bulk fill bits as the
existing algorithm was noticeably slow after applying the
"expand-to-row" helper to the DirectX renderer invalidation.
- All `til` import hierarchy is now handled in the parent `til.h` file
and not in the child files to prevent circular imports from happening.
We don't expect the import of any individual library file, only the
base one. So this should be OK for now.
## Validation Steps Performed
- Ran `cmatrix`, `cmatrix -u0`, and `cacafire` after changes were made.
- Made a bunch of ligatures with `Cascadia Code` in the Terminal
before/after the changes and confirmed they still ligate.
- Ran `dir` in Powershell and fixed the scrolling issues
- Clicked all over the place and dragged to make sure selection works.
- Checked retro terminal effect manually with Powershell.
## Summary of the Pull Request
When a pane is closed by a connection, we want to wait until the connection is actually `Closed` before we fire the actual `Closed` event. If the connection didn't close gracefully, there are scenarios where we want to print a message to the screen.
However, when a pane is closed by the UI, we don't really care to wait for the connection to be completely closed. We can just do it whenever. So I've moved that call to be on a background thread.
## PR Checklist
* [x] Closes#1996
* [x] I work here
* [ ] Tests added/passed
* [n/a] Requires documentation to be updated
## Detailed Description of the Pull Request / Additional comments
Previously we'd wait for the connection to close synchronously when closing tabs or panes. For misbehaving applications like `ssh.exe`, that could result in the `Close` needing to `WaitForSingleObject` _on the UI thread_. If the user closed the tab / pane either with a keybinding or with some other UI element, they don't really care to see the error message anymore. They just want the pane closed. So there's no need to wait for the actual connection to close - the app can just continue on with whatever it was doing.
## Validation Steps Performed
Messed around with closing tabs, panes, tabs with many panes, the entire window. Did this with keybindings, or by clicking on the 'x' on the tab, the 'x' on the window, or using middle-click.
I'm always scared of things like this, so there's a 50% chance this makes things horribly worse.
When the connection printed text immediately, synchronously, as part of
Start() it would cause terminal to deadlock. We should start the
connection outside of lock.
The ConptyConnection would do this when it failed to launch something
(trivial repro: `wt -- xyz`).
The TelnetConnection would do this all the time, because local loopback
telnet is fast and easy.
My basic idea was that `WM_CHAR` is just the better `WM_KEYDOWN`.
The latter fails to properly support common dead key sequences like in
#3516.
As such I added some logic to `Terminal::SendKeyEvent` to make it return
false if the pressed key represents a printable character.
This causes us to receive a character event with a (hopefully) correctly
composed code unit, which then gets sent to `Terminal::SendCharEvent`.
`Terminal::SendCharEvent` in turn had to be modified to support
potentially pressed modifier keys, since `Terminal::SendKeyEvent` isn't
doing that for us anymore.
Lastly `TerminalInput` had to be modified heavily to support character
events with modifier key states. In order to do so I merged its
`HandleKey` and `HandleChar` methods into a single one, that now handles
both cases.
Since key events will now contain character data and character events
key codes the decision logic in `TerminalInput::HandleKey` had to be
rewritten.
## PR Checklist
* [x] CLA signed
* [x] Tests added/passed
* [x] I've discussed this with core contributors already.
## Validation Steps Performed
* See #3516.
* I don't have any keyboard that generates surrogate characters. Due to
this I modified `TermControl::_SendPastedTextToConnection` to send the
data to `_terminal->SendCharEvent()` instead. I then pasted the test
string ""𐐌𐐜𐐬" and ensured that the new `TerminalInput::_SendChar`
method still correctly assembles surrogate pairs.
Closes#3516Closes#3554 (obsoleted by this PR)
Potentially impacts #391, which sounds like a duplicate of #3516
The terminal lock is really only for the terminal; since the renderer is
fully owned by the control, not the Terminal, and we'll only be
receiving swap chain events after we register them during
initialization, we don't need to lock before _or_ after firing off the
coroutine.
Fixes#5203.
## Summary of the Pull Request
`TrimWhitespace` is misleading. So it is now renamed as 'singleLine'. If true, it comes out as a single line! That makes more sense!
## PR Checklist
* [X] Closes#3824
## Validation Steps Performed
Attempted the keybinding with both settings (and none set).
Attempted mouse copy with and without shift.
## Summary of the Pull Request
As we've learned in #979, not all touchpads are created equal. Some of them have bad drivers that makes scrolling inactive windows not work. For whatever reason, these devices think the Terminal is all one giant inactive window, so we don't get the mouse wheel events through the XAML stack. We do however get the event as a `WM_MOUSEWHEEL` on those devices (a message we don't get on devices with normally functioning trackpads).
This PR attempts to take that `WM_MOUSEWHEEL` and manually dispatch it to the `TermControl`, so we can at least scroll the terminal content.
Unfortunately, this solution is not very general purpose. This only works to scroll controls that manually implement our own `IMouseWheelListener` interface. As we add more controls, we'll need to continue manually implementing this interface, until the underlying XAML Islands bug is fixed. **I don't love this**. I'd rather have a better solution, but it seems that we can't synthesize a more general-purpose `PointerWheeled` event that could get routed through the XAML tree as normal.
## References
* #2606 and microsoft/microsoft-ui-xaml#2101 - these bugs are also tracking a similar "inactive windows" / "scaled mouse events" issue in XAML
## PR Checklist
* [x] Closes#979
* [x] I work here
* [ ] Tests added/passed
* [n/a] Requires documentation to be updated
## Detailed Description of the Pull Request / Additional comments
I've also added a `til::point` conversion _to_ `winrt::Windows::Foundation::Point`, and some scaling operators for `point`
## Validation Steps Performed
* It works on my HP Spectre 2017 with a synaptics trackpad
- I also made sure to test that `tmux` works in panes on this laptop
* It works on my slaptop, and DOESN'T follow this hack codepath on this machine.
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
This PR will allow TSFInputControl to redraw its Canvas and TextBlock in response to when the Terminal cursor position updates. This will fix the issue where during Korean composition, the first symbol of the next composition will appear on top of the previous composed character. Since the Terminal Cursor updates a lot, I've added some checks to see if the TSFInputControl really needs to redraw. This will also decrease the number of actual redraws since we receive a bunch of `LayoutRequested` events when there's no difference between them.
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#4963
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
Startup, teardown, CJK IME gibberish testing, making sure the IME block shows up in the right place.
This commit rewrites a large swath of TermControl's initialization code.
* `TermControl` now _always_ has a `_terminal`; it will never be null
* Event registration for `_terminal` and any other available-at-init
fixtures has been moved into the constructor.
* Event handlers how more uniformly check `_closing` if they interact
with the _terminal.
* Swap chain attachment has been cleaned up and no longer uses a
coroutine when it's spawned from the UI thread.
* We have to register the renderer's swapchain change notification
handler after we set the swap chain, otherwise it'll call us back
when it initializes itself.
* `InitializeTerminal` now happens under the `_terminal`'s write lock
* Certain things that InitializeTerminal were calling themselves
attempted to take the lock. They no longer do so.
* TermControlAutomationPeer cannot take the read lock, because setting
the scrollbar's `Maximum` during `InitializeTerminal` will trigger
vivification of the automation peer tree; if it attempts to take the
lock it will deadlock during initialization.
* `BlinkCursor` was renamed to `CursorTimerTick` because it's the "Tick"
handler for the "CursorTimer".
* `DragDropHandler` was converted into a coroutine instead of just
_calling_ a coroutine.
Caveats:
Terminal may not have a `_buffer` until InitializeTerminal happens.
There's a nasty coupling between RenderTarget and TextBuffer that means
that we need to have a renderer before we have a buffer.
There's a second nasty coupling between RenderThread and Renderer: we
can't create a RenderThread during construction because it needs to be
given a renderer, and we can't create a Renderer during construction
because it needs a RenderThread. We don't want to kick off a thread
during construction.
Testing:
I wailed on this by opening and closing and resizing terminals and panes
and tabs, up to a hundred open tabs and one tab with 51 panes. I set one
tab to update the title as fast as it possibly could and tested
teardown, zoom, resize, mouse movement, etc. while this was all
happening.
Closes#4613.
This PR has evolved to encapsulate two related fixes that I can't really
untie anymore.
#2455 - Duplicating a tab that doesn't exist anymore
This was the bug I was originally fixing in #4429.
When the user tries to `duplicateTab` with a profile that doesn't exist
anymore (like might happen after a settings reload), don't crash.
As I was going about adding tests for this, got blocked by the fact that
the Terminal couldn't open _any_ panes while the `TerminalPage` was size
0x0. This had two theoretical solutions:
* Fake the `TerminalPage` into thinking it had a real size in the test -
probably possible, though I'm unsure how it would work in practice.
* Change `Pane`s to not require an `ActualWidth`, `ActualHeight` on
initialization.
Fortuately, the second option was something else that was already on my
backlog of bugs.
#4618 - `wt` command-line can't consistently parse more than one arg
Presently, the Terminal just arbitrarily dispatches a bunch of handlers
to try and handle all the commands provided on the commandline. That's
lead to a bunch of reports that not all the commands will always get
executed, nor will they all get executed in the same order.
This PR also changes the `TerminalPage` to be able to dispatch all the
commands sequentially, all at once in the startup. No longer will there
be a hot second where the commands seem to execute themselves in from of
the user - they'll all happen behind the scenes on startup.
This involved a couple other changes areound the `TerminalPage`
* I had to make sure that panes could be opened at a 0x0 size. Now they
use a star sizing based off the percentage of the parent they're
supposed to consume, so that when the parent _does_ get laid out,
they'll take the appropriate size of that parent.
* I had to do some math ahead of time to try and calculate what a
`SplitState::Automatic` would be evaluated as, despite the fact that
we don't actually know how big the pane will be.
* I had to ensure that `focus-tab` commands appropriately mark a single
tab as focused while we're in startup, without roundtripping to the
Dispatcher thread and back
## References
#4429 - the original PR for #2455#5047 - a follow-up task from discussion in #4429#4953 - a PR for making panes use star sizing, which was immensly
helpful for this PR.
## Detailed Description of the Pull Request / Additional comments
`CascadiaSettings::BuildSettings` can throw if the GUID doesn't exist.
This wraps those calls up with a try/catch.
It also adds a couple tests - a few `SettingsTests` for try/catching
this state. It also adds a XAML-y test in `TabTests` that creates a
`TerminalPage` and then performs som UI-like actions on it. This test
required a minor change to how we generate the new tab dropdown - in the
tests, `Application::Current()` is _not_ a `TerminalApp::App`, so it
doesn't have a `Logic()` to query. So wrap that in a try/catch as well.
While working on these tests, I found that we'd crash pretty agressively
for mysterious reasons if the TestHostApp became focused while the test
was running. This was due to a call in
`TSFInputControl::NotifyFocusEnter` that would callback to
`TSFInputControl::_layoutRequested`, which would crash on setting the
`MaxSize` of the canvas to a negative value. This PR includes a hotfix
for that bug as well.
## Validation Steps Performed
* Manual testing with a _lot_ of commands in a commandline
* run the tests
* Team tested in selfhost
Closes#2455Closes#4618
This commit rewrites selection handling at the TermControl layer.
Previously, we were keeping track of a number of redundant variables
that were easy to get out of sync.
The new selection model is as follows:
* A single left click will always begin a _pending_ selection operation
* A single left click will always clear a selection (#4477)
* A double left click will always begin a word selection
* A triple left click will always begin a line selection
* A selection will only truly start when the cursor moves a quarter of
the smallest dimension of a cell (usually its width) in any direction
_This eliminates the selection of a single cell on one click._
(#4282, #5082)
* We now keep track of whether the selection has been "copied", or
"updated" since it was last copied. If an endpoint moves, it is
updated. For copy-on-select, it is only copied if it's updated.
(#4740)
Because of this, we can stop tracking the position of the focus-raising
click, and whether it was part of click-drag operation. All clicks can
_become_ part of a click-drag operation if the user drags.
We can also eliminate the special handling of single cell selection at
the TerminalCore layer: since TermControl determines when to begin a
selection, TerminalCore no longer needs to know whether copy on select
is enabled _or_ whether the user has started and then backtracked over a
single cell. This is now implicit in TermControl.
Fixes#5082; Fixes#4477
This commit introduces a github action to check our spelling and fixes
the following misspelled words so that we come up green.
It also renames TfEditSes to TfEditSession, because Ses is not a word.
currently, excerpt, fallthrough, identified, occurred, propagate,
provided, rendered, resetting, separate, succeeded, successfully,
terminal, transferred, adheres, breaks, combining, preceded,
architecture, populated, previous, setter, visible, window, within,
appxmanifest, hyphen, control, offset, powerpoint, suppress, parsing,
prioritized, aforementioned, check in, build, filling, indices, layout,
mapping, trying, scroll, terabyte, vetoes, viewport, whose
## Summary of the Pull Request
Alt-Numpad# input would be escaping each numkey before sending it through. This would result in some weird behavior, for example, in powershell, where the first alt-numpad# would start digit argument and once the user releases alt, a character is sent through and digit argument would repeat that character X times. To resolve this, we simply need to ignore KeyDowns where Alt is held and a Numpad# is pressed.
Once Alt is released, we'll receive a character through `TSFInputControl`, not `TermControl::CharacterHandler`. It seems that the `CoreTextEditContext` in `TSFInputControl` intercepts the character before it gets to `TermControl`. TSF will then send the received character through as normal.
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#1401
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
Testing various combinations of Alt-Numpad# consistently sends through only one instance of the expected symbols.
Adds support for setting the cursor visibility in Terminal. Visibility
is a property entirely independent from whether the cursor is "on" or
not. The cursor blinker _should_ change the "IsOn" property. It was
actually changing the "Visible" property, which was incorrect. This PR
additionally corrects the naming of the method used by the cursor
blinker, and makes it do the right thing.
I added a pair of tests, one taken straight from conhost. In
copy-pasting that test, I took it a step further and implemented
`^[[?12h`, `^[[?12l`, which enables/disables cursor blinking, for the
`TerminalCore`. THIS DOES NOT ADD SUPPORT FOR DISABLING BLINKING IN THE
APP. Conpty doesn't emit the blinking on/off sequences quite yet, but
when it _does_, the Terminal will be ready.
## References
* I'd bet this conflicts with #2892
* This isn't a solution for #1379
* There shockingly isn't an issue for cursor blink state via conpty...?
## PR Checklist
* [x] Closes#3093
* [x] Closes#3499
* [x] Closes#4644
* [x] I work here
* [x] Tests added/passed
* [n/a] Requires documentation to be updated
## Summary of the Pull Request
Make TerminalControl synthesize mouse events and Terminal send them to
the TerminalInput's MouseInput module.
The implementation here takes significant inspiration from how we handle
KeyEvents.
## References
Closes#545 - VT Mouse Mode (Terminal)
References #376 - VT Mouse Mode (ConPty)
### TerminalControl
- `_TrySendMouseEvent` attempts to send a mouse event via TermInput.
Similar to `_TrySendKeyEvent`
- Use the above function to try and send the mouse event _before_
deciding to modify the selection
### TerminalApi
- Hookup (re)setting the various modes to handle VT Input
- Terminal is _always_ in VT Input mode (important for #4856)
### TerminalDispatch
- Hookup (re)setting the various modes to handle VT Input
### TerminalInput
- Convert the mouse input position from viewport position to buffer
position
- Then send it over to the MouseInput in TerminalInput to actually do it
(#4848)
## Validation Steps Performed
Tests should still pass.
If UseAcrylic is disabled, CTRL+SHIFT+SCROLL would enable it, without
having to change the setting in profile.json manually.
1. Set "useAcrylic" to false for the any profile in profile.json
2. Open terminal window for that profile.
3. CTRL+SHIFT+MouseScroll
Acrylic background opacity should change according to mouse scroll
## PR Checklist
* [x] CLA signed.
* [x] Tested manually
* [x] Updated documentation
Closes#661
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
I originally thought that setting `TSFInputControl::_editContext.InputPaneDisplayPolicy` to be Automatic would allow the InputPanel to show and hide automatically when `TSFInputControl` gains and loses focus. It doesn't seem to behave that way, so we'll show the InputPanel manually.
I'll show the panel during `PointerPressedHandler` and during `GotFocusHandler`. A user will have the on-screen keyboard pop up when getting focus, but if they close the keyboard, they can simply re-tap on the terminal to bring it back up.
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#3639
* [x] CLA signed.
* [x] Tests added/passed
## Validation Steps Performed
Played on my surfaces book with the on screen keyboard by closing/tapping on the terminal and using the search box.
<!-- Enter a brief description/summary of your PR here. What does it fix/what does it change/how was it tested (even manually, if necessary)? -->
## Summary of the Pull Request
Right clicking on a focused tab while Copy On Select is active currently copies any active selection. This is because `PointerReleasedHandler` doesn't check for the mouse button that was released.
During a mouse button release, only the left mouse button release should be doing anything.
<!-- Please review the items on the PR checklist before submitting-->
## PR Checklist
* [x] Closes#4740
* [x] CLA signed.
* [x] Tests added/passed
<!-- Describe how you validated the behavior. Add automated tests wherever possible, but list manual validation steps taken as well -->
## Validation Steps Performed
These are the scenarios I've tested. They're a combination of in focus/out of focus, Copy On Select on/off, left/right click pressed and their move and release variants.
From Out of Focus:
- Left Click = Focus
- Left Click Move = Focus + Selection
- Left Click Release
- CoS on = Copy
- CoS off = Nothing
- Shift Left Click = Focus
- Right Click
- Focus
- CoS on = Paste
- CoS off = Copy if Active Selection, Paste if not.
- Right Click Move = Nothing
- Right Click Release = Nothing
From In Focus
- Left Click = Selection if CoS off
- Left Click Move = Selection
- Left Click Release
- CoS on = Copy
- CoS off = Nothing
- Shift Left Click = Set Selection Anchor
- Right Click
- CoS on = Paste
- CoS off = Copy if Active Selection, Paste if not.
- Right Click Move = Nothing
- Right Click Release = Nothing
## Summary of the Pull Request
`GetTextForClipboard` already exists in the TextBuffer. It makes sense to use that for UIA as well. This changes the behavior or `GetText()` such that it does not remove leading/trailing whitespace anymore. That is more of an expected behavior.
## References
This also contributes to...
- #4509: UIA Box Selection
- #2447: UIA Signaling for Selection
- #1354: UIA support for Wide Glyphs
Now that the expansion occurs at before render-time, the selection anchors are an accurate representation of what is selected. We just need to move GetText to the TextBuffer. Then we can have those three issues just rely on code from the text buffer. This also means ConHost gets some of this stuff for free 😀
## Detailed Description of the Pull Request / Additional comments
- `TextBuffer::GetTextForClipboard()` --> `GetText()`
- `TextBuffer::GetText()` no longer requires GetForegroundColor/GetBackgroundColor. If either of these are not defined, we return a `TextAndColor` with only the `text` field populated.
- renamed a few parameters for copying text to the clipboard for clarity
- Updated `UiaTextRange::GetText()` to use `TextBuffer::GetText()`
## Validation Steps Performed
Manual tests for UIA using accessibility insights and Windows Terminal's copy action (w/ and w/out shift)
Added tests as well.
## Summary of the Pull Request
We're deref'ing a null `_terminal`. Don't do that. This is a _okay_ fix, mostly to stem the bleeding. @DHowett-MSFT's got a mind for a real fix to #4166, but this isn't it.
## PR Checklist
* [x] related to #4166
* [x] I work here
* [ ] Tests added/passed
* [n/a] Requires documentation to be updated
The Xaml input stack doesn't allow an application to suppress the "caret
browsing" dialog experience triggered when you press F7.
The official recommendation from the Xaml team is to catch F7 before we
hand it off.
This commit introduces a special F7 handler and an ad-hoc implementation of event bubbling.
Runtime classes implementing a custom IF7Listener interface are
considered during a modified focus parent walk to determine who can
handle F7 specifically.
If the recipient control handles F7, we suppress the message completely.
This event bubbler has some minor issues -- the search box will not be
able to receive F7 because its parent control implements the handler.
Since search is already mostly a text box, it doesn't _need_ special
caret browsing functionality as far as I can tell.
TermControl implements its OnF7Pressed handler by synthesizing a
keybindings event and an event to feed into Terminal Core directly.
It's not possible to create a synthetic KeyPressRoutedEvent; if it were,
I would have just popped one into the traditional input queue. :)
Fixes#638.
## Summary of the Pull Request
Korean IME was not working correctly due to way we were clearing the input buffer inside of `TSFInputControl`. We wanted to clear our input buffer and tell TSF to clear its input buffer as well when we receive a `CompositionCompleted` event. This works fine in some IME languages such as Chinese and Japanese. However, Korean IME composes characters differently in such a way where we can't tell TSF to clear their buffer during a `CompositionCompleted` event because it would clear the character that triggered the `CompositionCompleted` event in the first place.
The solution in this PR is to keep our `_inputBuffer` intact until the user presses <kbd>Enter</kbd> or <kbd>Esc</kbd>, in which case we clear our buffer and the TSF buffer. I've chosen these two keys because it seems to make sense to clear the buffer after text is sent to the terminal with <kbd>Enter</kbd>, and <kbd>Esc</kbd> usually means to cancel a current composition anyway.
This means we need to keep track of our last known "Composition Start Point", which is represented by `_activeTextStart`. Whenever we complete a composition, we'll send the portion of the input buffer between `_activeTextStart` and the end of the input buffer to the terminal. Then, we'll update `_activeTextStart` to be the end of the input buffer so that the next time we send text to the terminal, we'll only send the portion of our buffer that's "active".
## PR Checklist
* [x] Closes#4226
* [x] CLA signed
* [x] Tests added/passed
## Validation Steps Performed
Manual testing with Chinese, Japanese, and Korean IME.
## Summary of the Pull Request
Match conhost behavior and clear selection on viewport/font resize.
## PR Checklist
* [X] Closes#1165
## Validation Steps Performed
Retried attached bug repro steps
- When performing chunk selection, the expansion now occurs at the time
of the selection, not the rendering of the selection
- `GetSelectionRects()` was moved to the `TextBuffer` and is now shared
between ConHost and Windows Terminal
- Some of the selection variables were renamed for clarity
- Selection COORDs are now in the Text Buffer coordinate space
- Fixes an issue with Shift+Click after performing a Multi-Click
Selection
## References
This also contributes to...
- #4509: UIA Box Selection
- #2447: UIA Signaling for Selection
- #1354: UIA support for Wide Glyphs
Now that the expansion occurs at before render-time, the selection
anchors are an accurate representation of what is selected. We just need
to move `GetText` to the `TextBuffer`. Then we can have those three
issues just rely on code from the text buffer. This also means ConHost
gets some of this stuff for free 😀
### TextBuffer
- `GetTextRects` is the abstracted form of `GetSelectionRects`
- `_ExpandTextRow` is still needed to handle wide glyphs properly
### Terminal
- Rename...
- `_boxSelection` --> `_blockSelection` for consistency with ConHost
- `_selectionAnchor` --> `_selectionStart` for consistency with UIA
- `_endSelectionPosition` --> `_selectionEnd` for consistency with
UIA
- Selection anchors are in Text Buffer coordinates now
- Really rely on `SetSelectionEnd` to accomplish appropriate chunk
selection and shift+click actions
## Validation Steps Performed
- Shift+Click
- Multi-Click --> Shift+Click
- Chunk Selection at...
- top of buffer
- bottom of buffer
- random region in scrollback
Closes#4465Closes#4547
Defines the following automation properties for a Terminal Control:
- [**Orientation**](https://docs.microsoft.com/en-us/uwp/api/windows.ui.xaml.automation.peers.automationpeer.getorientationcore):
- The orientation of the control
- None --> Vertical
- [**Name**](https://docs.microsoft.com/en-us/uwp/api/windows.ui.xaml.automation.peers.automationpeer.getnamecore):
- The name as used by assistive technology and other Microsoft UI
Automation clients. Generally presented by automation clients as the
primary way to identify an element (along with the control type)
- "" --> <profile name>
- [**HelpText**](https://docs.microsoft.com/en-us/uwp/api/windows.ui.xaml.automation.peers.automationpeer.gethelptextcore):
- The help text. Generally presented by automation clients if
requested by the user. This would be something that you would normally
expect to appear from tooltips.
- "" --> <tab title>
- [**LiveSetting**](https://docs.microsoft.com/en-us/uwp/api/windows.ui.xaml.automation.peers.automationpeer.getlivesettingcore):
- reports the live setting notification behavior. A representation of
how assertive this control should be when content changes.
- none --> Polite
## Detailed Description of the Pull Request / Additional comments
ProfileName had to be added to the TerminalSettings (IControlSettings)
to pass that information along to the automation peer. In the rare event
that somebody purposefully decided to make their ProfileName empty, we
fallback to the tab title.
## Validation Steps Performed
Verified using Accessibility Insights and inspect.exe
This is are some examples of the information a general user can expect
to receive about a Terminal Control.
- Type: Terminal Control
- Name: Command Prompt
- Help Text (if requested): Command Prompt - ping bing.com
- Type: Terminal Control
- Name: Ubuntu
- Help Text (if requested): cazamor@PC-cazamor:/mnt/c/Users/cazamor$
Note, it is generally read by an automation client as follows:
"<type>, <name>"
References #2099 - Automation Properties for TerminalControl, Search Box
References #2142 - Localization
Closes#2142
This commit removes all of the custom UI initialization code in
TermControl and replaces it with a xaml file. Some dead or reundant code
was removed as part of this refactoring.
It also fixes two (quasi-related) issues:
* The search box, on first launch, was offset by the scrollbar even if
the scrollbar was invisible.
* The scrollbar state wasn't hot-reloadable.
## Summary of the Pull Request
Currently, when the user attempts to type using IME while the Search Box is focused, the input goes to the terminal instead. This is due to the fact that the `TSFInputControl` assumes it's in control whenever TermControl gets focus. So, it'll intercept IME input before the Search Box receives it. We simply need to modify `TermControl::GotFocus` to check if the SearchBox has focus. If it does, `TSFInputControl::NotifyFocusEnter` shouldn't be called.
As a small side fix, I've also disabled the terminal cursor blinking when the Search Box has focus.
Thinking a little further, if we have more features in the future that behave like search box (i.e. advanced tab switcher, or any other XAML controls that pop up) and require text input, we might need to create a sort of "AnyOtherTextControlInFocus" function to see if TSFInputControl should receive focus or not.
## PR Checklist
* [x] Closes#4434
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed
## Validation Steps Performed
Search works as expected with IME. Composition picker appears underneath Search Box when typing IME in the Search Box. Clicking outside of the Search Box still returns control to TSFInputControl/TermControl.
Terminal Cursor blinks when it has focus, and doesn't when the Search Box has focus.
## Summary of the Pull Request
We used to return multiple text ranges to represent one selection. We only support one selection at a time, so we should only return one range.
Additionally, I moved all TriggerSelection() calls to the renderer from Terminal to TermControl for consistency. This ensures we only call it _once_ when we make a change to our selection state.
## References
#2447 - helps polish Signaling for Selection
#4465 - This is more apparent as the problem holding back Signaling for Selection
## PR Checklist
* [x] Closes#4452
Tested using Accessibility Insights.
## Summary of the Pull Request
Currently, clicking on an unfocused terminal with a selection active will trigger `copyOnSelect`. This is because the check for `copyOnSelect` and copying to the clipboard is bound to when the Pointer is released. This works fine for when a user performs a click-drag selection, but it inadvertently also triggers when the user performs a single click on an unfocused terminal. We expect `copyOnSelect` to trigger only on the first time a selection is completed.
This PR will allow the user to single click on an unfocused terminal that has a selection active without triggering a copyOnSelect. It also ensures that any click-drag selection, whether it's on an unfocused or focused terminal, will trigger copyOnSelect.
## PR Checklist
* [x] Closes#4255
## Validation Steps Performed
Performed manual testing involving permutations of multiple panes, tabs, in focus, and out of focus.
Fixed inconsistent scrolling when using both touchscreen and precision
touchpad.
Scrolling jumpiness is caused by rounding errors. Instead of retrieving
the current scrolling value from `GetScrollOffset`, which is already
rounded in `int`, let the scrolling operation to operate on
`_scrollBar`'s `Value` directly (which uses `double`) in
`TermControl.cpp`.
TermControl now also respects WHEEL_DELTA, which it was previously
ignoring, to determine how many scroll wheel detents were used.
Testing scrolling on the following scenario manually:
- nonscrollable terminal (e.g. the window is large enough to contain the
current buffer).
- scrolling to the topmost and bottom-most.
- scrolling TUI apps such as `nano` and `more` in WSL.
- after clearing the terminal, both in cmd and WSL.
... has the same behavior between using touchscreen or precision trackpad
and regular mouse wheel.
Closes#4554 (original pull request)
Closes#1066Closes#4542
## Summary of the Pull Request
The issue seems to be how `SwapChainScaleChanged` gets fired and attempts to tell the renderer
to `UpdateDPI` when the renderer is gone. So, as a quick bandaid, we'll put a quick check to only do the thing if the renderer is alive.
## PR Checklist
* [x] Closes#4539
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed
## Validation Steps Performed
Held my new tab button for about thirty seconds then held the close tab button until all tabs closed without a crash.
## Summary of the Pull Request
This PR tries to address some of the weird interactions with pointer pressed events when the Terminal isn't in focus. Here's the four things that have changed as part of this PR;
1. This PR will allow the user to be able to make a selection with a click-drag without having to first perform a single click on a tab/pane to bring it to focus.
2. Another weird bug that's fixed in this PR is where trying to make a selection on an unfocused tab when it already has a selection active will simply extend the existing selection instead of making a new one.
3. Not related to the issue that his PR closes: a right click will now focus the tab/pane.
I've made sure that we still have the existing functionality where a single click on an unfocused tab/pane does not make a single-cell selection and just focuses the tab/pane.
## PR Checklist
* [x] Closes#4282
* [x] CLA signed. If not, go over [here](https://cla.opensource.microsoft.com/microsoft/Terminal) and sign the CLA
* [x] Tests added/passed
## Validation Steps Performed
Played around with all sorts of selection when in-focus and out of focus with multiple panes and tabs.
Unit tests still pass as well.
## Summary of the Pull Request
This will collect some user choices related to profiles and tab settings to help us understand if and how we should change the in-built defaults.
## PR Checklist
* [x] Closes#3855
* [x] I work here.
* [x] Manual test only.
* [x] Meh, no doc update.
* [x] Am core contributor.
## Detailed Description of the Pull Request / Additional comments
The following data is collected with examples of the types of questions we intend to answer:
1. What is the name of the executable attached to the PTY? (What shells are popular? Should we focus our testing on them? Are there any common ones we are blind to that we should know about?)
- "Microsoft.Windows.Terminal.Connection" {e912fe7b-eeb6-52a5-c628-abe388e5f792}
- "ConPtyConnected" event
- "SessionGuid" value = WT_SESSION
- "Client" value = Name of EXE
2. Is Acrylic used on a tab? And with what opacity? (Do people really want acrylic? Should it be default? What opacity is most pleasing in our context?)
- "Microsoft.Windows.Terminal.App" {24a1622f-7da7-5c77-3303-d850bd1ab2ed}
- "TabInformation" event
- "EventVer" value is now 1u
- "UseAcrylic" value is now TRUE/FALSE on the setting choice
- "TintOpacity" value is now Float on the setting choice
3. What font are people choosing? (Do people move away from Cascadia Code? Which ones are the most popular for us to validate when updating the renderer?)
- "Microsoft.Windows.Terminal.App" {24a1622f-7da7-5c77-3303-d850bd1ab2ed}
- "TabInformation" event
- "FontFace" value is now string font from settings
4. What keybindings do people choose to customize (Add or Remove)? (Are there extremely common keys that folks bind or unbind that we should have adjusted by default in a fresh install?)
- "Microsoft.Windows.Terminal.App" {24a1622f-7da7-5c77-3303-d850bd1ab2ed}
- "CustomKeybindings" event
- "Keybindings" value is the entire JSON segment that describes the user keybindings from `settings.json`.
5. Do people change their default profile from the PowerShell one we set? If so, to what? (Should we not set PowerShell as the default? Should we adjust the ranking of our dynamic generators to favor the most popular ones to bubble to the top?)
- "Microsoft.Windows.Terminal.App" {24a1622f-7da7-5c77-3303-d850bd1ab2ed}
- "CustomDefaultProfile" event
- "DefaultProfile" value is the GUID of the chosen profile
## Validation Steps Performed
1. Implemented the events
2. Launched the ETL channel viewer
3. Triggered the events
4. Saw the data come out
